COBOL Trends
in [Cobol]
|
From: Michael Wojcik on 16 Sep 2009 13:12 robertwessel2(a)yahoo.com wrote: > On Sep 14, 11:46 am, Howard Brazee <how...(a)brazee.net> wrote: >> Is the Z-80 still the top selling computer chip in the world? It's >> real cheap, and plenty fast enough to run a stop light. > > No. The Z-80 *chip* sells in tiny quantities. It would be far, far, > far too expensive to use a discrete microprocessor in almost any > embedded design suited for that class of CPU. > > OTOH, Z-80 *cores*, as embedded into other chips (including freely > available soft cores that can but programmed into FPGAs), are fairly > popular. In terms of shipped volume, at least the cores in the PIC, > AVR, and 8051 families are likely an order of magnitude higher in > volume than Z-80 cores. ARMs are likely higher volumes than Z-80s as > well. FWIW, reliable figures are notoriously difficult to get a hold > of. And since the Z80 patents have expired and the architecture's passed into public domain, there are a number of companies manufacturing Z80-compatible cores. (Rabbit's cores seem to be relatively popular right now, for example.) Numbers *are* hard to come by. There's the ESD survey,[1] but it's voluntary and asks about current projects, not existing stuff that's still being cranked out without redesign. The ESD survey suggests that 32-bit cores are growing very fast in embedded applications that don't actually need them (ie, for things like redesigns of working applications that get along fine with one or two 8- or 16-bit controllers). At some point we'll turn the corner and 32-bit cores will outsell 8-bit cores, even in applications that aren't doing anything much more than showing the clock and driving the occasional pin. I don't think that's happened yet, but at the moment I don't have any reliable references from the past few years. The 2006 ESD survey does show AVR as the most popular 8-bit family, with PIC 18 in second place and PIC 14/16 in fourth, so it agrees with what Robert hypothesized above. [1] ftp://ftp.embedded.com/pub/ESD%20SubscribSurvey/2006%20ESD%20Market%20Study.pdf -- Michael Wojcik Micro Focus Rhetoric & Writing, Michigan State University
From: Michael Wojcik on 16 Sep 2009 12:40 Pete Dashwood wrote: > Michael Wojcik wrote: >> Pete Dashwood wrote: >>> Certainly the paradigm that COBOL represented has been replaced in >>> Client/Server processing. As client/server (networking, the >>> Internet, etc.) is where MOST of the computer use in the world is >>> occurring, it is fair to say that COBOL is being replaced. >> Client/server computing is not "most of the computer use in the >> world". > > Yes it is. No, it isn't. We can do this all day. >> Most of the computers sold in recent years are embedded >> systems (and a majority of those are 8-bitters). The type of computer >> with the most users, worldwide, are mobile phones - a quintessential >> peer-to-peer application. If compute cycles is our metric, most >> computer use is in scientific number crunching. > > That's just silly. We are talking computers used by people interacting with > software. People interact with embedded computers all the time. If you're only talking about general-purpose computers, then perhaps you should have qualified your initial claim. > A smart washing machine or toaster is not part of this discussion > and "compute cycles" was NOT the metric for my statement. >> If we restrict "computer use" to mean people using general-purpose >> computers for a task they explicitly initiate, the dominant >> application is email. That does involve clients and servers, but it's >> hardly an interesting example of "client/server processing". > > So you want to move my statement to include "interesting" now? > > "interesting" is a completely subjective perspective. You think email is not > interesting as a technology; I find it fascinating, and dealing with it on > web based applications is challenging and satisfying. I didn't say email wasn't interesting; I said it wasn't an interesting example of client/server computing. It existed long before anyone thought to coin the term "client/server computing", because even though the usual email architecture (MUA / MTA) involves clients and servers, there's nothing striking or innovative about that. It's the arrangement that's obvious to any experienced practitioner. "Client/server computing" became a buzzword only when that architecture began appearing in applications where it wasn't the most obvious arrangement - applications that could easily be monolithic. Email isn't one of those applications. Consequently, email doesn't mark any sort of paradigm shift to client/server computing, and so it doesn't support any claims about the importance of client/server computing as an idea. >> Personally, I doubt that even in classic business back-office >> processing a majority of transactions are client/server in any >> interesting sense. But that's much harder to measure, since many large >> organizations don't even know what they run internally on a daily >> basis. (That's why there's a market for application portfolio analysis >> tools.) > > The people who work on the shop floor know what they run. Rarely. Look at some help desk logs. Most computer users have no idea what software packages they're using. >> This is the same error we see from Web 2.0 pundits, New Media >> enthusiasts, "long tail" proponents and the like - they ignore the >> sectors of the industry that don't fit their models, and consequently >> mistake the innovations of the vanguard for a revolution of the >> masses. > > I'm not convinced. I didn't expect you would be. This error is pervasive among pundits and analysts; few people bother to look at any actual data. > Talk to anyone under thirty and ask them what a computer is. I know people under thirty who have reasonable definitions for "computer", and I don't really care how people who don't know what they're talking about define it. > AND they all have one and have been using it all their lives. You completely > overlooked the fact that the internet is taking around 2 billion web page > hits a day, much of this off social _NETWORKS_ (my emphasis). No I haven't. It's still a minority function. There are more than 2 billion online financial transactions a *second*. > Pedantically there may be more computing cycles consumed by car engine > management systems, but that has no bearing on COBOL, which is what I was > discussing. And this is precisely the problem with the Web 2.0 pundits (Tim "Two-point" O'Reilly and the like) and similar commentators. They ignore everything in the industry that doesn't interest them. It's fine to say something like "client/server computing seems to be the dominant paradigm in general-purpose computing", or even better support that with some data. It's fine to say something like "social networking sites are seeing a lot of traffic [even better, be explicit about what 'a lot' is relative to], and that's interesting". But to make unqualified claims about what "computing" is, with handwaving generalities, often leads to just the sort of ridiculous nonsense that Tim and friends keep spouting. It's intellectually lazy and misleading. -- Michael Wojcik Micro Focus Rhetoric & Writing, Michigan State University
From: robertwessel2 on 16 Sep 2009 15:14 On Sep 16, 12:12 pm, Michael Wojcik <mwoj...(a)newsguy.com> wrote: > The ESD survey suggests that 32-bit cores are growing very fast in > embedded applications that don't actually need them (ie, for things > like redesigns of working applications that get along fine with one or > two 8- or 16-bit controllers). At some point we'll turn the corner and > 32-bit cores will outsell 8-bit cores, even in applications that > aren't doing anything much more than showing the clock and driving the > occasional pin. I don't think that's happened yet, but at the moment I > don't have any reliable references from the past few years. > > The 2006 ESD survey does show AVR as the most popular 8-bit family, > with PIC 18 in second place and PIC 14/16 in fourth, so it agrees with > what Robert hypothesized above. One thing to remember about the EDS survey's is that they end up being about *designs*, and mostly ignore shipping volumes. Very few designs are made with four bit MCUs, since the cost differential to an eight- bitter is so small, but they tend to ship disproportionate numbers (where a few pennies per unit savings is worthwhile) EDS reports less than 1% for designs, but 4-bit shipments are probably around 5% of the cores. Sixteen bitters in a different boat - in terms of units they've been severely squeezed between the eight bitters (which are cheaper), and the much more capable 32 bitters. They are, however typically easier to program than the eight bitters, and tend to be easier to integrate in low volume systems, so they get a disproportionate number of design wins (there are also some 16 bit DSPs, which form the bottom of that market). The EDS number for 32 bits are also clearly about designs, since ease of design is paramount for the large majority of projects - and if you need more than what you can put in an eight-bitter, picking up an ARM with a (relatively) fantastic tool chain is very tempting - but the EDS (design) numbers are 55% for 32 bitters, while *shipping* units is probably around 10%.
From: tlmfru on 16 Sep 2009 22:20 Those of us that have been around for a while know what a plastic term "client-server" is. When it first came in I wrote to a number of I/T publications urging them to require that if an author used the term, s/he should define it as well. (None of 'em did, more's the pity.) One author gave as an example of a client-server success a company that sorted its warehouse picking slips to bin number sequence! Obviously Paul, the person to whom you're replying, needs to specify his meaning. PL Michael Wojcik <mwojcik(a)newsguy.com> wrote in message news:h8r72j21d46(a)news2.newsguy.com... > The only way to support the claim that client/server processing > represents a majority of compute cycles is to broaden the definition > to absurdity and call things like MPP and cellphone traffic > "client/server". At that point the term is no longer useful. You might > as well claim the embedded CPU in a USB keyboard is a client and the > PC it's attached to is a server, and say that's client/server as well. > > -- > Michael Wojcik > Micro Focus > Rhetoric & Writing, Michigan State University
From: John Reagan on 17 Sep 2009 10:56
"Howard Brazee" <howard(a)brazee.net> wrote in message news:hpssa597gjhriua7nk620rsmp6e9j3rs32(a)4ax.com... > Is the Z-80 still the top selling computer chip in the world? It's > real cheap, and plenty fast enough to run a stop light. > I don't think so. Looking around at the McCain controllers (no relation to the Senator), the older model 170 controllers use a 6802. Subsequent controllers have two CPUs. A MC68360 and a MC68302. And the most recent models use a Freescale PowerQUICC (essentially a PowerPC). Their software runs a full IP stack and lots of other things you'd never fit on an 8-bit processor. http://www.mccain-inc.com/traffic/category/controllers.html John |